26
A Novel Marking Mechanism for Packet Video Delivery over DiffServ Networks Haidong Wang, Guizhong Liu School of Electronic and Information Engineering, Xi’an Jiaotong University, Xi’an China

A novel marking mechanism for packet video delivery over diff serv networks

Embed Size (px)

Citation preview

Page 1: A novel marking mechanism for packet video delivery over diff serv networks

A Novel Marking Mechanism for Packet Video Delivery over DiffServ Networks

Haidong Wang, Guizhong Liu

School of Electronic and Information Engineering,Xi’an Jiaotong University, Xi’an China

Page 2: A novel marking mechanism for packet video delivery over diff serv networks

Page 2

Introduction

In recent years, there have been more and more streaming multimedia applications over the Internet. As a result, packet video transmissions on the Internet have increased sharply.

However, the current IP networks, which only provide best-effort service, fail to supply any quality of services (QoS).

Page 3: A novel marking mechanism for packet video delivery over diff serv networks

Page 3

Introduction

IETF has defined DiffServ service models.

The marker is used to mark the packets as green, yellow or red.

The core router discards the packets based on specified rules when the traffic congestion of the networks occurs

Marker(red,yellow or green)

Core Router(Congestion

Dropper)

Page 4: A novel marking mechanism for packet video delivery over diff serv networks

Page 4

Introduction

One of the key issues in the DiffServ network architectures is how to efficiently mark the packets of an IP traffic stream with the three colors (red, yellow or green), which stand for three drop levels in a single AF(Assured Forwarding) class.

Page 5: A novel marking mechanism for packet video delivery over diff serv networks

Page 5

Introduction

IETF proposed a Two Rate Three Color Marker (TRTCM), which meters an IP packet stream and marks its packets based on two rates, namely, PIR (Peak Information Rate) and CIR (Committed Information Rate).

However, there exists the relative importance in packet video streams, e.g., an I frame packet is more important than a P or B frame packet, which is not taken into account in TRTCM.

Page 6: A novel marking mechanism for packet video delivery over diff serv networks

Page 6

Introduction

Ref [1] presented an Enhanced Token Bucket Three Color Marker (ETBTCM) with consideration of both the network condition and the relative importance of video packets.

[1] C.-H. Ke, C.-K. Shieh, W.-S. Hwang, and A. Ziviani, “A two marker system for improved mpeg video delivery in a diffserv network,” IEEE Communications Letters, vol. 9, no. 4, pp. 381 – 383, 2005.

Main idea:

sufficient tokens -> all packets are green.

insufficient tokens -> red for B frames and yellow for I or P frames

Page 7: A novel marking mechanism for packet video delivery over diff serv networks

Page 7

Introduction

Ref[2] presented an Improved Two Rate Three Color Marker (ITRTCM).

[2]L. Chen, G. Liu, and F. Zhao, “An improved marking mechanism for real-time video over diffserv networks,” vol. 4810 LNCS, 2007, pp. 510– 519.

Main idea:

This method employ a number of thresholds for dividing the network status into several grades , and 16 priority grades denoted the video packets importance.

Page 8: A novel marking mechanism for packet video delivery over diff serv networks

Page 8

Introduction

ETBTCM and ITRTCM could improve the end-to-end video quality of the packet video by use of the relative importance of the video packets.

But the marking mechanisms are inaccurate when the token count is close to the thresholds.

For example, A packet of the B frames gets to the marker at

the moment that the token count is beyond zero, and it is green. At the next moment, a packet of the P frames gets to the marker and the token count is less than zero unfortunately, and it is yellow.

Page 9: A novel marking mechanism for packet video delivery over diff serv networks

Page 9

Introduction

In this paper, we develop a new marker called Priority-Aware Two Rate Three Color Marker (PATRTCM)

PATRTCM retains the token buckets and its associated parameters, but introduces three marking probabilities, which are calculated in terms of the relative importance of the video packet and the traffic conditions in the network.

According to the probabilities, a video packet is marked red, yellow or green.

Page 10: A novel marking mechanism for packet video delivery over diff serv networks

Page 10

The Priority Classification Scheme

There were defined Q importance grades for video packets, ranging from 1 to Q.

The basic idea of the classification is that the I frame packets are of the highest priority, the B frames packets are of the lowest priority, and the P frames packets are classified according to their decoding order.

Page 11: A novel marking mechanism for packet video delivery over diff serv networks

Page 11

PATRTCM

The PATRTCM retains the C and P buckets, also retains the CIR, CBS, PIR and PBS parameters.

PATRTCM defines three probabilities, i.e., Pred, Pyellow and Pgreen.

Page 12: A novel marking mechanism for packet video delivery over diff serv networks

Page 12

PATRTCM

Assume that a given video packet of size S bytes get to the marker at time t, prio represents the priority of the packet Tc(t) represents the token count in C at time t Tp(t) represents the token count in P at time t Tc(0)=CBS and Tp(0)=PBS.

The network status is divided into 3 different grades roughly by use of the three conditions, Tp(t)-S<0, Tp(t)-S>0 and Tc(t)-S<0, Tc(t)-S>0.

Page 13: A novel marking mechanism for packet video delivery over diff serv networks

Page 13

PATRTCM

The fundamental principle of PATRTCM is: Pred + Pyellow + Pgreen = 1. For the light loading, the more sufficient tokens

are available in the buckets, the larger Pgreen is; and the higher priority the packet is, the larger its Pgreen is.

For the middle loading, the more sufficient tokens are available in the buckets, the larger Pyellow is; and the higher priority the packet is, the larger its Pyellow is.

For the heavy loading, Pred=1

Page 14: A novel marking mechanism for packet video delivery over diff serv networks

Page 14

PATRTCM

(1) Tp(t)-S<0 (the heavy loading) The packet is marked red without regard

to its priority, i.e., Pred=1, Pyellow=0, Pgreen=0.

(2) Tp(t)-S>0 and Tc(t)-S<0 (the middle loading)

Let thP=PBS / Q

th0=(prio- 1)×thP

th1=prio×thP

th2=(prio+ 1)×thP

Page 15: A novel marking mechanism for packet video delivery over diff serv networks

Page 15

PATRTCM

Tp(t) - S≤th0 th0<Tp(t) - S≤th1 th1<Tp(t) - S≤th2 Tp(t) - S>th2

Pred=1 Pred =1 - Pyellow Pred =0 Pred =0

Pyellow=0Pyellow =[Tp(t) - S- th0 ] / (th1 -

th0)Pyellow =1 - Pgreen Pyellow =0

Pgreen=0 Pgreen=0Pgreen =[Tp(t) - S

- th1 ] / (th2 -th1)

Pgreen=1

(2) Tp(t)-S>0 and Tc(t)-S<0 (the middle loading)

Page 16: A novel marking mechanism for packet video delivery over diff serv networks

Page 16

PATRTCM

(3) Tc(t)-S>0 (the light loading) If prio<Q, the packet is green, i.e.,

Pred=0, Pyellow=0, Pgreen=1 If prio=Q

Let thC=CBS / 2

Pred=0

Pyellow=1 - Pgreen

Pgreen=[Tc(t) - S] / thC

Page 17: A novel marking mechanism for packet video delivery over diff serv networks

Page 17

Simulation

The simulations are performed using the NS2 simulator

Page 18: A novel marking mechanism for packet video delivery over diff serv networks

Page 18

Simulation

The simulation network consists of a video sender/receiver, 9 exponential distribution on-off background traffics senders/receivers, an ingress router, a core router and an egress router.

We suppose that there be a bottleneck link between the core router and the egress router, and all the other links be 10Mbps each.

Page 19: A novel marking mechanism for packet video delivery over diff serv networks

Page 19

Simulation

The test video sequences are Highway with CIF format with 2000 frames, and Foreman with CIF format with 300 frames.

The sequences are encoded into the standard H.264 bitstreams with a mean bit rate of 500 kbps.

The rate of each competing exponential on-off traffic is 500 kpbs, the burst time is 500 ms and the idle time is 100 ms.

Page 20: A novel marking mechanism for packet video delivery over diff serv networks

Page 20

Simulation

The bottleneck bandwidth (denoted BW) is changed in order to obtain 6 different levels of payload.

Page 21: A novel marking mechanism for packet video delivery over diff serv networks

Page 21

Simulation

The core router implements the Weighted Random Early Detection (WRED) for the Active Queue Management (AQM)

There are 3 parameters in WRED:

Page 22: A novel marking mechanism for packet video delivery over diff serv networks

Page 22

Simulation

We compare the 4 schemes in the experiment: TRTCM, ETBTCM, ITRTCM and PATRTCM.

Page 23: A novel marking mechanism for packet video delivery over diff serv networks

Page 23

Results (Highway)

Page 24: A novel marking mechanism for packet video delivery over diff serv networks

Page 24

Results (Foreman)

Page 25: A novel marking mechanism for packet video delivery over diff serv networks

Page 25

Conclusion

Simulation results show that PATRTCM can improve the end-to-end video quality over DiffServ Networks.

However, we leave the bandwidth fairness of TCP flows out of consideration currently, which is a future research topic we will focus on.

Page 26: A novel marking mechanism for packet video delivery over diff serv networks

Page 26

Thanks !